MRI image segmentation using machine learning networks and level set approaches

<span>The segmented brain tissues from magnetic resonance images (MRI) always pose substantive challenges to the clinical researcher community, especially while making precise estimation of such tissues. In the recent years, advancements in deep learning techniques, more specifically in fully convolution neural networks (FCN) have yielded path breaking results in segmenting brain tumour tissues with pin-point accuracy and precision, much to the relief of clinical physicians and researchers alike. A new hybrid deep learning architecture combining SegNet and U-Net techniques to segment brain tissue is proposed here. Here, a skip connection of the concerned U-Net network was suitably explored. The results indicated optimal multi-scale information generated from the SegNet, which was further exploited to obtain precise tissue boundaries from the brain images. Further, in order to ensure that the segmentation method performed better in conjunction with precisely delineated contours, the output is incorporated as the level set layer in the deep learning network. The proposed method primarily focused on analysing brain tumor segmentation (BraTS) 2017 and BraTS 2018, dedicated datasets dealing with MRI brain tumour. The results clearly indicate better performance in segmenting brain tumours than existing ones.</span>

Quality assessment of anatomical MRI images from Generative Adversarial Networks: human assessment and image quality metrics

Background: Generative Adversarial Networks (GANs) can synthesize brain images from image or noise input. So far, the gold standard for assessing the quality of the generated images has been human expert ratings. However, due to limitations of human assessment in terms of cost, scalability, and the limited sensitivity of the human eye to more subtle statistical relationships, a more automated approach towards evaluating GANs is required. New method: We investigated to what extent visual quality can be assessed using image quality metrics and we used group analysis and spatial independent components analysis to verify that the GAN reproduces multivariate statistical relationships found in real data. Reference human data was obtained by recruiting neuroimaging experts to assess real Magnetic Resonance (MR) images and images generated by a Wasserstein GAN. Image quality was manipulated by exporting images at different stages of GAN training. Results: Experts were sensitive to changes in image quality as evidenced by ratings and reaction times, and the generated images reproduced group effects (age, gender) and spatial correlations moderately well. We also surveyed a number of image quality metrics which consistently failed to fully reproduce human data. While the metrics Structural Similarity Index Measure (SSIM) and Naturalness Image Quality Evaluator (NIQE) showed good overall agreement with human assessment for lower-quality images (i.e. images from early stages of GAN training), only a Deep Quality Assessment (QA) model trained on human ratings was sensitive to the subtle differences between higher-quality images. Conclusions: We recommend a combination of group analyses, spatial correlation analyses, and both distortion metrics (SSIM, NIQE) and perceptual models (Deep QA) for a comprehensive evaluation and comparison of brain images produced by GANs.

A Novel 2-Phase U-net Algorithm Combined with Optimal Mass Transportation for 3D Brain Tumor Detection and Segmentation

Utilizing the optimal mass transportation (OMT) technique to convert an irregular 3D brain image into a cube, a required input format for the U-net algorithm, is a brand new idea for medical imaging research. We develop a cubic volume-measure-preserving OMT (V-OMT) model for the implementation of this conversion. The contrast-enhanced histogram equalization grayscale of fluid attenuated inversion recovery (FLAIR) in a brain image creates the corresponding density function. We then propose an effective two-phase U-net algorithm combined with the V-OMT algorithm for training and validation. First, we use the U-net and V-OMT algorithms to precisely predict the whole tumor (WT) region. Second, we expand this predicted WT region with dilation and create a smooth function by convoluting the step-like function associated with the WT region in the brain image with a 5×5×5 blur tensor. Then, a new V-OMT algorithm with mesh refinement is constructed to allow the U-net algorithm to effectively train Net1--Net3 models. Finally, we propose ensemble voting postprocessing to validate the final labels of brain images. We randomly choose 1000 and 251 brain samples from theBraTS 2021 training dataset, which contains 1251 samples, for training and validation, respectively. The Dice scores of the WT, tumor core (TC) and enhanced tumor (ET) regions for validation computed by Net1--Net3 were 0.93705, 0.90617 and 0.87470, respectively. A significant improvement in brain tumor detection and segmentation with higher accuracy is achieved.

Crossed Aphasia after Intracranial Hematoma in the Right Temporoparietal Lobe: A Case Report

Purpose: Aphasia in a dextral after right hemisphere injury is called crossed aphasia (CA). We are reporting a first case of transformation of motor aphasia to conduction aphasia after right hemisphere intracerebral hemorrhage (ICH) associated with arteriovenous malformation (AVM) with literature reviews.Methods: A case of a man in his 30s with CA following right hemisphere ICH in the temporal-parietal lobe associated with AVM was reviewed. We analyzed his brain images, initial linguistic characteristics, and changes in aphasia for 8 weeks of follow-up.Results: The initial Paradise Korean Western Aphasia Battery Revised (PK-WAB-R) was evaluated as aphasia quotient (AQ) 72, 64%ile; post evaluation was evaluated AQ 95, 98.9%ile after 8 weeks. The post-test repetition score was the patient score range, which can be attributed to impairment in phonological short-term memory. The patient is diagnosed anomalous CA based on Alexander et al., and we could predict that the language ability originates from both hemispheres based on Nagaraja et al. Considering the appearance of Gerstmann syndrome at the beginning of the onset, we could expect that the function of the dominant parietal lobe is partially crossed as well.Conclusion: Changes in aphasia were reported throughout the initial stage to the end of speech therapy. It is also important to note that literature review of Korean studies was analyzed in this study. It will be necessary to conduct a cognitive test in the early stage of onset to understand the language problems of crossed conduction aphasia to know the characteristics of the cognitive process.

Effects of COVID-19 on the human central olfactory system: a natural pre-post experiment

Reduced olfactory function is the symptom with the highest prevalence in COVID-19 with nearly 70% of individuals with COVID-19 experiencing partial or total loss of their sense of smell at some point during the disease. The exact cause is not known but beyond peripheral damage, studies have demonstrated insults to both the olfactory bulb and central olfactory brain areas. However, these studies often lack both baseline pre-COVID-19 assessments and a control group and could therefore simply reflect preexisting risk factors. Right before the COVID-19 outbreak, we completed an olfactory focused study including structural MR brain images and a full clinical olfactory test. Opportunistically, we invited participants back one year later, including 9 participants who had experienced mild to medium COVID-19 (C19+) and 12 that had not (C19-), thereby creating a pre-post controlled natural experiment with a control group. Despite C19+ participants reporting subjective olfactory dysfunction, few showed signs of objectively altered function one year later. Critically, all but one individual in the C19+ group had reduced olfactory bulb volume with an average volume reduction of 14.3%, but this did not amount to a significant between group difference compared to the control group (2.3% reduction) using inference statistics. No morphological differences in cerebral olfactory areas were found but we found stronger functional connectivity between olfactory brain areas in the C19+ croup at the post measure. Taken together, these data suggest that COVID-19 might cause a long-term reduction in olfactory bulb volume but with no discernible differences in cerebral olfactory regions.

TReC: Transferred ResNet and CBAM for Detecting Brain Diseases

Early diagnosis of pathological brains leads to early interventions in brain diseases, which may help control the illness conditions, prolong the life of patients, and even cure them. Therefore, the classification of brain diseases is a challenging but helpful task. However, it is hard to collect brain images, and the superabundance of images is also a great challenge for computing resources. This study proposes a new approach named TReC: Transferred Residual Networks (ResNet)-Convolutional Block Attention Module (CBAM), a specific model for small-scale samples, to detect brain diseases based on MRI. At first, the ResNet model, which is pre-trained on the ImageNet dataset, serves as initialization. Subsequently, a simple attention mechanism named CBAM is introduced and added into every ResNet residual block. At the same time, the fully connected (FC) layers of the ResNet are replaced with new FC layers, which meet the goal of classification. Finally, all the parameters of our model, such as the ResNet, the CBAM, and new FC layers, are retrained. The effectiveness of the proposed model is evaluated on brain magnetic resonance (MR) datasets for multi-class and two-class tasks. Compared with other state-of-the-art models, our model reaches the best performance for two-class and multi-class tasks on brain diseases.

The Impact of Resampling and Denoising Deep Learning Algorithms on Radiomics in Brain Metastases MRI

Background: Magnetic resonance imaging (MRI) is predominant in the therapeutic management of cancer patients, unfortunately, patients have to wait a long time to get an appointment for examination. Therefore, new MRI devices include deep-learning (DL) solutions to save acquisition time. However, the impact of these algorithms on intensity and texture parameters has been poorly studied. The aim of this study was to evaluate the impact of resampling and denoising DL models on radiomics. Methods: Resampling and denoising DL model was developed on 14,243 T1 brain images from 1.5T-MRI. Radiomics were extracted from 40 brain metastases from 11 patients (2049 images). A total of 104 texture features of DL images were compared to original images with paired t-test, Pearson correlation and concordance-correlation-coefficient (CCC). Results: When two times shorter image acquisition shows strong disparities with the originals concerning the radiomics, with significant differences and loss of correlation of 79.81% and 48.08%, respectively. Interestingly, DL models restore textures with 46.15% of unstable parameters and 25.96% of low CCC and without difference for the first-order intensity parameters. Conclusions: Resampling and denoising DL models reconstruct low resolution and noised MRI images acquired quickly into high quality images. While fast MRI acquisition loses most of the radiomic features, DL models restore these parameters.

Fractal analysis as a method of morphometric study of linear anatomical objects: modified Caliper method

The purpose of the study was to develop an original modification of the Caliper method of image fractal analysis to determine the fractal dimension of linear anatomical objects. To develop the method, the linear contour of the outer surface of the cerebral cortex was chosen as the object of study. Magnetic resonance brain images in coronal projection were used. The original modification of the Caliper method includes image analysis using Adobe Photoshop CS5 software or its analogues. The linear contour of the studied object is selected, followed by stepwise smoothing of the contour with different smoothing radius. At the 1st stage of fractal analysis smoothing is not applied, at the 2nd stage the smoothing radius is 2 pixels, the 3rd – 4 pixels, the 4th – 8 pixels, the 5th – 16 pixels. At each stage, the contour length in pixels is measured (P). The size of the fractal measurement unit (G) at the 1st stage of fractal analysis is 1 pixel, the 2nd stage – 2 pixels, the 3rd stage – 4 pixels, the 4th stage – 8 pixels, the 5th stage – 16 pixels. The contour smoothing radius, the size of the fractal measurement units and the number of stages of fractal analysis can be changed depending on the characteristics of the studied structure, size, scale and image resolution. Based on the values of the perimeter and the size of the fractal measurement units, the number of fractal measurement units covering the studied object (N) is calculated: N=P/G. The fractal dimension value is calculated based on the N and G values. The modification of the Caliper method described in this paper is automatized and does not require much time required for manual calculation. In addition, compared to the classic Caliper method, this modification is more accurate because the measurement is performed automatically. The main limitation of the developed modification is the ability to determine the fractal dimension of only closed contours of studied structures or closed linear structures, because this method involves determining the length of the closed perimeter of the selected image area. The modified Caliper method of image fractal analysis described in this paper can be used in morphology and other fields of medicine for fractal analysis of linear objects: external and internal linear contours of different anatomical structures (cerebellum, cerebral hemispheres) and pathological foci (tumors, foci of necrosis, fibrosis, etc.).